 |
INTRODUCTION |
Hydrophobicity is the primary driving force for polypeptide
integration into membranes. Moreover, the precise sequence and/or length of a hydrophobic transmembrane sequence is a critical factor in
regulating transmembrane topology (1). Yet, accumulating evidence also
suggests a role for charged residues, primarily in determining the
orientation of hydrophobic transmembrane segments in membranes (2, 3).
In general, the more positively charged end of the transmembrane
sequence is located on the cytoplasmic side of either the endoplasmic
reticulum (ER)1 or bacterial
plasma membrane (4, 5). In Escherichia coli, positively
charged residues have been shown to play an active role in
post-translational (mostly Sec-independent) integration into membranes
(6, 7). Furthermore, it appears that positive charges regulate topology
by blocking protein translocation (7). In yeast, it appears that both
negatively and positively charged residues can contribute to the
overall charge distribution to govern orientation (8). Recently, other
non-hydrophobic effectors of membrane topology have been reported,
including both negatively charged residues in E. coli (9)
and the folded state of the amino terminus (3).
In eukaryotes, cell-free assay systems have revealed numerous details
about the translocation process and also provided indications of
considerable regulatory complexity (10). In addition to sequences that
regulate topology, cell-free systems have been used to identify sequences that regulate translocation at the ER. A block in
translocation has been described for plasminogen activator inhibitor-2
that alters the interaction of the nascent polypeptide with the signal recognition particle and with the translocation machinery in the ER
membrane. The translocation block results in both secreted and
cytoplasmic forms of the protein (11). There is also evidence that
P-glycoprotein, a membrane protein predicted to span the membrane 12 times, adopts at least two different membrane topologies (12, 13).
Cell-free assays were also used to demonstrate that the protein ductin
adopts two different topologies at the ER membrane (14). It was later
determined that both forms are physiologically relevant. One form of
ductin is a subunit of the vacuolar H+-ATPase, and the
other is found in gap junctions. Dual topology has also been
demonstrated for the L protein of hepatitis B virus. In this case the
molecule appears to be initially integrated into the ER membrane in a
single orientation and then a subset of the polypeptides are
post-translationally translocated across the ER membrane to generate
the alternate orientation (15).
In contrast to these sequences, which appear to determine the final
orientation of a transmembrane segment in the membrane, there is also
evidence for sequence-specific regulation of polypeptide integration
into the ER membrane (10). Integration into membranes may be regulated
for some sequences that are not sufficiently hydrophobic to simply
partition into the lipid bilayer. For example, correct membrane
assembly of some integral membrane proteins involves the integration of
multiple transmembrane segments that individually are unable to
integrate into the lipid bilayer (10). Co- and post-translational
translocation mechanisms involving the first two transmembrane
sequences have been implicated in correct assembly of the cystic
fibrosis transmembrane conductance regulator (16). Furthermore, an
unusual transmembrane sequence that appears to be regulated for
integration into the ER membrane was identified as part of
transmembrane sequence 7 from P-glycoprotein (17).
Finally, an element termed a stop-transfer effector (STE) sequence has
been identified that mediates membrane integration of an otherwise
secreted hydrophobic domain (18). The two known STE sequences were
identified in the transmembrane form of murine IgM (19) and in the
prion-related protein, PrP (20, 21).
Many immunoglobulins are synthesized in both secreted and transmembrane
forms. Over most of the length of the molecules, the two forms are
identical, but the transmembrane forms contain an additional sequence
at the carboxyl terminus of the polypeptide. This sequence encodes a
hydrophilic sequence of approximately 20 amino acids, followed by the
hydrophobic transmembrane domain of the protein. Differential splicing
of the primary transcript that adds another exon to the mRNA
accounts for the change in protein sequence (22, 23). In
vitro, a sequence of amino acids including the hydrophilic
sequence amino-terminal of the murine IgM transmembrane sequence was
demonstrated to have STE activity when positioned at either the amino
terminus or in the middle of a secreted protein (19).
Although the mechanism of STE-mediated membrane integration is poorly
characterized, recent data suggest that it may be fundamentally different than membrane integration mediated by hydrophobicity alone.
For example, unlike conventional topology-determining sequences, the
function of the PrP STE sequence depends on both the composition of the
cell-free translation extract (20) and involves a nascent polypeptide-encoded pause in translocation (24). Moreover, recent data
demonstrate that when inappropriate membrane integration by the PrP STE
is provoked in transgenic animals, neural degeneration results that is
very similar to that seen in Gerstmann-Straussler Scheinker syndrome, a
correlation that suggests that STE regulation is important in human
disease (25). However, little is known about the specific properties of
STE sequences that trigger integration into membranes because the two
known STE sequences show no sequence similarity. Thus, important
residues in these STE sequences must be identified experimentally.
Here we have used a mutagenesis strategy along with photocross-linking
to decipher the signal within the IgM STE that leads to membrane
integration. Membrane integration is shown to depend on the unusually
large number of negatively charged residues within the STE.
Surprisingly, the precise position of the charges is unimportant, and
the contribution of each of the negatively charged residues to
integration of the polypeptide into the membrane is additive. It
appears that STE recognition occurs within the translocon, since
functional and non-functional STE sequences both photocross-link the
translocon-associated protein TRAM. More importantly, functional STE
sequences photocross-link to two proteins, not previously identified as
translocon components that may recognize STE elements and effect their function.
| |
EXPERIMENTAL PROCEDURES |
Plasmid Constructs--
Plasmids were constructed by introducing
restriction sites into the wild-type sequence by site-directed
mutagenesis, and then all further substitutions were made by inserting
oligonucleotides into these sites within the plasmids. The positions of
the relevant restriction sites relative to the coding sequence is
indicated below the sequence in Fig. 1A. All of
the plasmids were sequenced using standard techniques. All plasmids
except C1 were inserted following the SP6 RNA polymerase promoter and
5'-untranslated leader in pSPUTK. In C1, the plasmid is identical to
pSPUTK except that an Ala at the 
3 position of the UTK leader is
substituted with a Cys. The coding sequences of the different plasmids
differ only in the regions indicated as the hydrophobic or the STE
region in Fig. 1. Unless specified otherwise the hydrophobic region was derived from the hydrophobic core of the preprolactin signal sequence by deleting the last seven amino acids and replacing an amino-terminal arginine with a cysteine (19). In WT, the hydrophobic region contains
the hydrophobic core sequence from IgM (19). An additional set of
plasmids was generated by replacing the hydrophobic region with the
regulated portion of transmembrane sequence 7 from human P-glycoprotein
(17). The signal cleavage region is common to all of the constructs and
is arbitrarily defined as the 9 amino acids after the hydrophobic
region. This sequence contains a cryptic signal peptide cleavage site.
Thus the signal cleavage region is the first 9 amino acids of the Pt
domain (amino acids 58-199 of prolactin) that was used as a reporter.
The control plasmids encode related non-hydrophobic residues and
positive charges (C1 and C2) or a single Met (C3) in place of the STE
region in the other constructs. Five plasmids were constructed in which
the STE and hydrophobic region are positioned internally in the
polypeptide between two reporter domains. These constructs included
STEs A2, A7, A16, C1, and C2. Another three plasmids were constructed
in which the hydrophobic region was replaced by a segment
(NGGLQPAFAIIFSKTTGVFTRIDDPETKRQ) from transmembrane sequence 7 from
P-glycoprotein (17). In these plasmids, the first reporter sequence
(globin amino acids 1-107) is preceded with a secretory signal peptide
(from preprolactin), and thus the entire polypeptide is translocated
into the interior of the microsome unless the STE stops translocation.
The construction of these plasmids was via cassette mutagenesis using
the restriction sites in Fig. 1 but starting with a plasmid encoding
the globin and Pt domains reported previously (19). Construction and
characterization of the plasmid encoding the WT STE and hydrophobic
transmembrane region of murine membrane IgM fused to amino acids
58-199 of bovine preprolactin were reported previously (19). The
sequence MVTER overlined in Fig. 1 indicates extra amino
acids added to the WT STE during cloning. Full construction details for
each plasmid are available from the authors.
Analysis of Topology and Membrane Integration--
Plasmids were
transcribed in vitro using SP6 polymerase, and then
polypeptides were synthesized from unpurified transcription products
using either a rabbit reticulocyte lysate translation system or a wheat
germ extract supplemented with canine pancreatic microsomes. Prior to
proteolysis, the canine pancreatic microsomes were separated from
untargeted molecules in the reticulocyte lysate translation reaction by
gel filtration chromatography on Sepharose CL-2B resin. The excluded
fractions containing microsomes and microsome-associated proteins (150 µl) were pooled, divided into 3 aliquots of 45 µl, and proteinase K
was added to two of the samples at 0.03 mg/ml for 15 min at 0 °C.
Proteolysis was terminated by adding trichloroacetic acid to a final
concentration of 16% (w/v). Precipitated proteins were washed with
ethanol/ether (1:1, v/v), and resuspended in SDS-PAGE loading buffer.
After electrophoresis, radioactivity was recorded from the dried
polyacrylamide gel using a PhosphorImager (Molecular Dynamics).
To create a reagent that would modify cysteine residues in the fusion
protein, but would not cross the ER membrane, we reacted N-(
-maleimidobutyryloxy) succinimide ester (GMBS, Pierce)
with spermine to synthesize N-(-maleimidobutyryloxy)
spermine amide (GMBSA). The resulting compound contains a reactive
maleimide attached to one of the four amino groups of spermine. For
this reaction, spermine was dissolved in 50% (v/v) acetonitrile/water to a final concentration of 180 mM. 150 µl of 0.5 M triethylamine (pH 7.5) was added to 400 µl of the
spermine solution, and the resulting solution was covered with parafilm
to prevent evaporation. GMBS was dissolved in 100% acetonitrile to a
final concentration of 180 mM, and 200 µl was added to
the spermine solution in 20-µl aliquots at 1-min intervals through a
small hole in the parafilm cover. The reaction was allowed to proceed
for an additional 40 min at room temperature and then the reaction
mixture containing GMBSA was reduced to approximately 200 µl in a speedvac.
For labeling membrane proteins from translation reactions, microsomes
were isolated from a 10-µl translation reaction using a Sepharose
CL-2B gel exclusion column equilibrated in 10 mM Tris-HCl (pH 7.5), 50 mM KCl, 2 mM MgCl2.
Protease inhibitors were added to the excluded volume (90 µl) to
final concentrations of 1 µg/ml chymostatin, antipain, leupeptin,
pepstatin, and 2 µg/ml aprotinin. The excluded fraction containing
the membranes was divided into two equal fractions, adjusted to 30 mM Tris-HCl (pH 7.5), and one was labeled by adding 4.5 µl of GMBSA and incubating at 24 °C for 1 h. The reactions
were quenched by adding dithiothreitol to 20 mM and
incubating for 5 min at 24 °C. The membranes were then pelleted from
the reactions by centrifugation at 50,000 × g for 15 min at 4 °C through a 100 µl 0.5 M sucrose cushion in a 200-µl Airfuge tube. To assess the GMBSA reactivity (topology) of
integrated proteins, the reactions were processed as above except that
they were adjusted to 1 M urea, 0.2 M sodium
carbonate (pH 11.5) before centrifugation. Membrane pellets were
resuspended directly in SDS-PAGE loading buffer and separated on a
15-cm 16% Laemmli polyacrylamide gel.
Photocross-linking--
Photoreactive probes were incorporated
into nascent polypeptides by adding
N
-(5-azido-2-nitrobenzoyl)-Lys-tRNA to
in vitro translations containing wheat germ extract,
truncated mRNA, the signal recognition particle, and microsomes,
and samples were then photolyzed and analyzed as described previously
(26, 27).
| |
RESULTS |
Characterization of the STE Sequence--
The additional exon
expressed in the transmembrane form of murine IgM includes both the
hydrophobic transmembrane sequence and a 21-residue putative STE
sequence (19). Similar hydrophilic sequences, containing 5-7
negatively charged residues, are found directly amino-terminal of the
transmembrane sequence of many immunoglobulins from species as diverse
as human and teleost fish. However, both the positions and the relative
proportion of glutamic and aspartic acid residues vary. In mouse, the
transmembrane form of IgM contains one aspartic and seven glutamic
residues, whereas immunoglobulin 2a (A15) contains five aspartic and
two glutamic acid residues (Fig. 1).
View larger version (35K):
[in this window]
[in a new window]
|
Fig. 1.
A, amino acid sequences (in single
letter code) of the STE, H (hydrophobic), and
C (signal cleavage) regions of the constructs analyzed. The
approximate positions of the restriction endonuclease sites in the
corresponding cDNA sequence used to construct the mutants are
indicated below the sequence. B, properties of
the mutant STE sequences. In WT the bar above the
sequence MVTER indicates extra amino acids added to the STE sequence
during cloning (19). The residue indicated with the 1 is the
first amino acid of the IgM STE. The M at the amino terminus
of each sequence is the initiator methionine. Dots indicate
residues unchanged from the WT sequence. An extra Ala residue inserted
in A1-A5 is indicated by a superscript. The
length of each hydrophobic region is the number of uncharged amino
acids between the last charge in the STE and the first residue of the
signal cleavage region. The fraction of polypeptides inserted into the
ER membrane when the STE was positioned at the amino terminus of the Pt
reporter is indicated by % TM; the standard error for that
determination is given by the number in brackets to the
right. a, the hydrophobic region in the control
molecule with the WT STE is the authentic IgM hydrophobic sequence
TTASTFIVLFLLSLFYSTTVTLF. *, molecules for which % TM was
not determined when located at the amino terminus of the
reporter.
|
|
There is no precedent for negatively charged amino acids controlling
polypeptide integration into membranes (28). However, our preliminary
deletion analysis of the IgM STE suggested that the Glu residues in the
sequence might contribute to STE function (data not shown). Therefore,
plasmids encoding a series of fusion proteins were constructed to
determine if the negatively charged residues within the IgM STE
sequence can contribute to polypeptide integration into ER membranes
(Fig. 1). Each fusion protein contains the topogenic element being
tested for STE activity and one or more protein domains used as
reporter sequences.
Regulation of polypeptide integration into membranes and the topologies
of the fusion proteins containing each of the various STE sequences
were assayed by proteolysis, carbonate extraction, and chemical
labeling, after synthesis in reticulocyte lysate or wheat germ extract
containing canine pancreatic microsomes. Finally, photocross-linking
was used to compare the environments occupied by functional and
non-functional STE sequences at the ER membrane.
The hydrophobic region used as a potential transmembrane domain was
derived from the preprolactin signal sequence because it is unlikely
that a hydrophobic sequence from a cleaved secretory signal sequence
would contain information other than hydrophobicity that specifies
membrane topology or integration. To generate the hydrophobic region
from the preprolactin signal sequence, the last seven amino acids of
the sequence were deleted (including the signal cleavage site), and an
amino-terminal Arg was replaced with Cys. Deletion of the signal
cleavage site greatly reduced signal peptide cleavage when the sequence
is positioned internally within a secreted protein (less than 5% of
molecules are cleaved, data not shown). Although all of the
polypeptides contained the same hydrophobic region, the overall length
of the hydrophobic portion of the potential transmembrane segment is
also determined by the distribution of charged residues in the STE and
the uncharged residues in the reporter that immediately precedes the
putative STE sequence. Substitution of the negative charges within the STE with uncharged residues can also change the length of the potential
transmembrane segment (Fig. 1).
Stop-Transfer Activity at an Internal Location--
To examine
assembly of a conventional type I transmembrane protein, chimeric
proteins were constructed containing a secretory signal sequence
followed by a reporter sequence derived from globin (Gt, amino acids
1-107 of chimpanzee globin) and another reporter sequence from
preprolactin (Pt, amino acids 58-199 of bovine preprolactin). The
sequences used as reporters have been shown previously to contain no
cryptic topogenic elements (29). To test the stop-transfer activity of
the various STE sequences and hydrophobic region combinations, they
were positioned between the Gt and Pt reporter sequences (Fig.
2).
View larger version (21K):
[in this window]
[in a new window]
|
Fig. 2.
Negative charges are required for
STE-mediated integration of marginally hydrophobic sequences into the
ER membrane. A, sequences were tested for STE activity with
hydrophobic sequences of 15, 17, or 23 residues (hatched,
black, and open bars, respectively). Nucleotides
encoding the test sequences (containing 1 or 4 Glu residues and the
indicated hydrophobic sequence) were positioned between sequences
coding for the Gt and Pt reporters. Proteolysis of the Pt reporter was
used to identify molecules with transmembrane topology after
transcription of the plasmid in vitro and translation of the
transcription products in reticulocyte lysate with added canine
pancreatic microsomes. The STE sequences analyzed were C1, C2, A2 (one
negative charge), and A16 and A7 (4 negative charges), from
left to right. B, when positioned within a
secreted reporter protein, molecules containing a functional STE (4 Glu
residues, open box labeled 4) and a marginally
hydrophobic region (15-17 residues) integrate as type I (amino
terminus lumenal) proteins with the Gt domain inside and the Pt domain
outside of the microsome. Non-functional STE mutants (1 Glu residue,
open box labeled 1) are unable to stop
translocation and are released into the ER lumen. The secretory signal
sequence at the amino terminus of the Gt domain is cleaved from both
molecules and is not shown. C, the negative charges in the
STE sequence are required for integration into membranes of a segment
of the transmembrane segment 7 sequence from P-glycoprotein.
Upward arrowheads indicate the protease-protected globin
fragment diagnostic of transmembrane molecules. The intensity of this
band is much less than for the full-length molecules because it
contains only 4 of the 8 methionine residues in the full-length
protein. Correction for the number of methionines revealed that 11, 18, and 23% of the molecules were transmembrane for STEs A3, A7, and A13,
respectively. Downward arrowheads indicate secreted
signal-cleaved molecules. Comparison of total reaction products
(lanes 1, 5, and 9) with membrane-bound molecules
indicates that the amino-terminal signal peptide is cleaved when the
molecules are translocated. The number of negative charges and the name
of each STE are indicated above the lanes. The STE sequences
are as shown in Fig. 1 except that the last three amino acids (NLW)
were replaced with a single amino acid (Val). The migration positions
of molecular mass markers (in kDa) are shown to the right of
the panel.
|
|
As expected, the number of molecules that adopted a transmembrane
topology depended on the length of the uncharged region of the
potential transmembrane sequence. When this region included 23 amino
acids (e.g. STE A2), approximately 85% of the molecules were transmembrane irrespective of whether or not a functional STE
sequence was included in the molecule (Fig. 2). When other non-functional STE sequences were added to shorter potential
transmembrane sequences of 15 (C1) or 17 (C2) residues, approximately
15 and 40% of the molecules were transmembrane, respectively (Fig. 2). In contrast, addition of an STE sequence containing 4 Glu residues to
the amino terminus of the same potential transmembrane regions (A16 and
A7) increased the transmembrane molecules to approximately 40 and 70%,
respectively (Fig. 2A). Thus, for potential transmembrane sequences of limited hydrophobicity, STE sequences with as few as 4 Glu
residues significantly increased the number of molecules that adopted a
transmembrane topology.
The hydrophobic region used in these molecules was derived from the
preprolactin signal sequence but was extensively modified and therefore
served only as a hydrophobic sequence that does not on its own remain
transmembrane. Although this sequence is not derived from an authentic
transmembrane domain, when the uncharged region of the sequence is
extended to 23 amino acids, it becomes transmembrane independent
of the STE sequence (Fig. 2A), as expected from previous
observations (1).
To determine whether or not the STE would stop translocation of an
authentic transmembrane domain, it was necessary to fuse the STE
sequence to a transmembrane sequence that is not sufficiently hydrophobic to mediate STE-independent membrane integration. The few
sequences that have been identified that are marginally hydrophobic yet
known to be transmembrane are generally found in multispanning membrane
proteins. Of these, transmembrane domain 7 from human P-glycoprotein
has been shown to stop poorly when expressed in reporter constructs.
Therefore, we examined the effect of the STE sequences on the topology
of fusion proteins containing a region of transmembrane domain 7 that
was previously shown to become transmembrane only when flanked by
hydrophobic transmembrane sequences (17). Thus, the coding regions of
these plasmids are very similar to those used in Fig. 2A
except that the hydrophobic region was replaced with the sequence
NGGLQPAFAIIFSKTTGVFTRIDDPETKRQ. The STE sequences differ from A3, A7,
and A13 (Fig. 1) only in the last three amino acids (NLW), which were
replaced with a single residue (Val) in the new constructs.
The results of control experiments indicated that in the absence of an
STE sequence the fusion protein was translocated across the ER membrane
(data not shown). Addition of the STE sequence with 6 negative charges
clearly increased the number of molecules adopting a transmembrane
topology (Fig. 2C). Whereas the STE with only 2 negative
charges had a barely detectable effect on topology, the STE with 4 negative charges resulted in an intermediate number of transmembrane
molecules. Thus, we conclude that the IgM STE can stop translocation of
a regulated transmembrane domain.
Detailed examination of the residues required for STE activity was not
possible using these constructs because substitution of charged
residues for uncharged residues eventually converts the mechanism of
integration from one dependent on the STE to one entirely due to
hydrophobicity. It was not possible to avoid this complication by
substitution of the negatively charged amino acids with sequences
containing positively charged residues because the positively charged
residues abolished STE function (data not shown). When the hydrophobic
region was positioned at the amino terminus of a reporter protein, it
was cleaved by signal peptidase and removed from the reporter protein
if it failed to be recognized as a bona fide transmembrane
domain. Cleavage occurred even when the hydrophobicity was increased by
the various charged residue substitutions. For this reason we examined
the function of the hydrophobic region alone and fused behind the
various STE sequences when positioned at the amino terminus of the Pt
reporter domain.
Stop-Transfer Activity at an Amino-terminal Location--
When the
hydrophobic region was positioned at the amino terminus of a reporter
domain, it functioned primarily as a secretory signal sequence, and
approximately 80% of the polypeptides were cleaved by signal
peptidase. The remaining molecules were not cleaved by signal
peptidase, presumably because the authentic signal cleavage site was
deleted, and cleavage at the cryptic site is less efficient. Since
relatively long uncharged sequences may still function as cleaved
secretory signal sequences instead of integrated signal-anchor
sequences (28), it was possible to substitute the charged residues in
the STE with uncharged amino acids and assay STE activity by the extent
of conversion of the signal sequence (translocation) to a signal-anchor
sequence (integration). Thus, at either an amino-terminal or internal
location, an STE and hydrophobic region become transmembrane if the STE
is functional but pass through the ER membrane if the STE is
nonfunctional or absent.
Mutants were analyzed to evaluate the importance of various aspects of
the STE sequence as follows: total length (e.g. A14 versus A13); addition of a positively charged residue
(e.g. A13 versus A12); length of hydrophobic
sequence (substitution of Glu with non-charged residues at the carboxyl
end of the STE sequence increases the length of the hydrophobic region,
e.g. A11 versus A13); and/or replacement of
single or multiple residues as indicated in Fig. 1B. To
compare the Glu-rich IgM STE with a similar sequence containing Asp
residues, we also examined the putative STE from murine immunoglobulin
2a (A15). As secreted controls we used a sequence containing primarily
hydrophilic residues or a single methionine (STEs C1 and C3,
respectively). As a transmembrane control we used a previously
characterized construct containing both the wild-type STE and the
hydrophobic transmembrane sequence of IgM fused to the Pt reporter
(19). This fusion protein also contains the additional amino acids
MVTER at the extreme amino terminus (Fig. 1B).
In the constructs assayed here, if a functional STE sequence is fused
to the amino terminus of the hydrophobic region, the chimera is
converted from a signal peptide into a signal-anchor sequence, and the
Pt domain is anchored as a transmembrane protein with a mixture of type
I and type II topologies (Fig. 3). Thus, depending on the extent of signal peptide cleavage, any particular fusion protein can have up to four fates when it interacts with membranes in vitro (Fig. 3B). If the STE does not
halt translocation (i.e. is non-functional), the molecule is
secreted and the signal peptide is or is not cleaved (Fig. 3A,
upward and downward arrowheads, respectively). If
translocation is halted by the STE, the transmembrane proteins can
adopt type I (protease-sensitive) or type II (partial protease
protection, Fig. 3A, dots) orientations.
View larger version (25K):
[in this window]
[in a new window]
|
Fig. 3.
Proteolysis assays used to determine membrane
topology. A, secreted molecules (arrowheads) and
the lumenal portion of the type II transmembrane molecules
(dots) are protected from added protease (Pk)
unless the membrane is solubilized with detergent (Det). For
type I (amino terminus, lumenal) transmembrane proteins, the Pt
reporter is protease-sensitive. In type II proteins, only the
cytoplasmically localized STE sequence at the amino terminus is
digested by the protease. Fully secreted molecules, both signal
sequence-uncleaved (downward pointing arrowheads) and signal
sequence-cleaved (upward pointing arrowheads), are protected
from protease. Since membrane fractions were separated from the
reactions by gel filtration chromatography prior to analysis by
proteolysis, non-targeted molecules are not seen. The number of
negative charges and the name of each STE are indicated
above the lanes. The migration positions of molecular mass
markers are indicated in kDa to the right of the
panels. B, when positioned at the amino terminus of the
reporter protein, molecules containing a functional STE integrate as
type I (amino terminus lumenal) or type II (amino terminus cytoplasmic)
proteins. Non-functional STE mutants are unable to stop translocation
and may or may not be signal-cleaved when they are released into the ER
lumen (arrowhead labeled S).
|
|
All of the fusion proteins were examined for translocation by
translation in reticulocyte lysate containing canine pancreatic microsomes, followed by separation of targeted from non-targeted molecules by gel filtration chromatography and proteolysis of the
former. Translocated molecules (containing a non-functional STE) are
protected from proteolysis because they are inside the microsome. Most
of the translocated molecules were also cleaved by signal peptidase. In
contrast, transmembrane molecules (containing a functional STE) are
partially or completely digested by the added protease.
When polypeptide disposition was assessed as detailed below, a direct
correlation (slope = 9.1, r = 0.96) was observed
between the number of glutamic residues and the stop-transfer
efficiency of the STE sequence (Fig.
4A). The numerical data used
to generate Fig. 4A and the error associated with each data
point are presented in Fig. 1B. This direct correlation was
particularly surprising given the variety of differences between the
STE sequences tested. For example, polypeptides with the A6 and A7 STEs
both contain four negative charges but differ at eight other positions,
and yet both polypeptides are about 50% transmembrane. In contrast, the polypeptides with STEs A13 and A7 (six and four negative charges, respectively) vary at only these two positions, yet the stop-transfer efficiencies of these STEs (approximately 75 and 50%, respectively) are significantly different.
View larger version (18K):
[in this window]
[in a new window]
|
Fig. 4.
STE function correlates with the number of
glutamic acid residues. A, the fraction of transmembrane
molecules depends on the number of negatively charged residues in the
STE sequences. B, signal cleavage correlates inversely with
the number of negative charges in the STE sequence. C, the
fraction of molecules in a type II transmembrane orientation correlates
with the number of negatively charged residues in the STE sequence.
D, the fraction of type I transmembrane molecules is
relatively constant with increased negative charge in the STE.
E, the fraction of transmembrane molecules does not
correlate with the length of the hydrophobic region. The hydrophobic
region is arbitrarily defined as the sequence between the predicted
cleavage site and the nearest charged residue in the STE sequence.
F, the fraction of integrated molecules (measured by
extraction with 0.2 M sodium carbonate (pH 11.5) and 1 M urea) correlates with the number of negative charges in
the STE. The values plotted in each graph are averages from
at least three independent experiments (except F, which is
from a single experiment done in triplicate) and were corrected for
methionine content.
|
|
Remarkably, the putative STE from Ig2a (A15) containing 5 Asp and
only 2 Glu residues resulted in a quantitatively similar number of
transmembrane molecules as that obtained for polypeptides with the A13
STE that contains 6 Glu residues (78 and 75%, respectively). Thus, we
conclude that STE function is effected by the negative charges, and it
is not important whether the charges result from Asp or Glu residues.
Most of the translocated molecules are signal-cleaved. Hence, it is not
surprising that the extent of signal peptide cleavage clearly
correlated inversely (slope = -9.5, r = 0.96) with
the number of glutamic residues in the STE sequence (Fig. 4B). However, lack of signal cleavage was not in itself
sufficient to halt translocation, since some uncleaved molecules were
secreted (Fig. 3A, downward pointing arrowheads).
Furthermore, lack of signal peptide cleavage for the 15-residue
hydrophobic region used here did not prevent translocation when located
in the interior of a secreted protein (Fig. 2A).
Transmembrane Orientation Determined by Proteolysis--
In
addition to determining the fraction of the molecules that are
transmembrane, proteolysis data were also used to determine the
transmembrane orientation of the molecules. Molecules with type II
topology are easily measured since these molecules are partially
protected from added protease. As expected there is a strong
correlation between the fraction of molecules with type II
transmembrane topology and the number of negative charges in the STE
sequences (slope 6.8, r = 0.87, Fig. 4C).
This correlation can be appreciated by visual inspection of the
PhosphorImager data since the band that results from proteolytic
removal of the short cytoplasmic amino terminus increases in intensity
(dots, Fig. 3).
Unlike type II molecules, polypeptides with type I topology are almost
completely digested by added protease (the lumenal and transmembrane
domain together are predicted to be less than 50 amino acids long and
contain only a single methionine). The fraction of type I molecules was
calculated from the PhosphorImager data after correcting for the number
of methionine residues in each species by first adding together the
signals from the bands resulting from uncleaved but secreted
polypeptides and the type II transmembrane polypeptides in the plus
protease lane (downward pointing arrowheads and
dots, respectively) and subtracting this sum from the signal
from the upper band (uncleaved molecules) in the minus protease lane
(Fig. 4D). The fraction of molecules adopting a type I
topology varied between just less than 20 and approximately 50%. With
the exception of molecules with STE A15, the correlation between the
fraction of the polypeptides that adopt type I topology and the number
of negative charges in the STE (slope = 2.3, r = 0.35) is marginal. If the data point due to STE A15 (7 negative
charges) is excluded, the calculated slope of is 1.7 with
r = 0.30.
Because of the uncertainty involved in determining the fraction of
molecules with type I topology by this calculation, we confirmed these
results using chemical labeling. Very similar results were obtained
when type I molecules were quantified after labeling with GMBSA, a
membrane impermeant modifier of cysteine (see below).
Although most of the STE-dependent increase in
transmembrane molecules was due to an increase in the number of
molecules with type II topology, STE-dependent triggering
of transmembrane topology is not dependent on which side of the ER
membrane the STE ultimately resides. For example, approximately half of
the roughly 75% transmembrane molecules with STE A12 or A13 adopted
the type I topology.
Examination of the topology adopted by each of the mutants revealed
that individual glutamic residues, the single lysine and the length of
the hydrophobic sequence each, by itself, has only limited influence on
the topology adopted. This result contrasts directly with data obtained
using conventional signal-anchor sequences where topology is directly
related to the distribution of charged residues surrounding the
hydrophobic segment.
Hydrophobicity Does Not Correlate with Stop-Transfer
Activity--
Unexpectedly, no correlation was observed between the
number of uninterrupted uncharged residues in the hydrophobic sequence and stop-transfer activity (Fig. 4E). Moreover, the A2 STE
and hydrophobic region functioned as a relatively efficient secretory signal peptide (71% fully translocated) even though it contains a 23 hydrophobic residue sequence followed by a sequence (CHTSSLPTP) that is
not particularly hydrophilic. Only the His and Pro in this latter
sequence are rare in transmembrane sequences (30). There was also no
correlation when stop-transfer activity was compared with total
calculated hydrophobicity of the hydrophobic sequence in each molecule
(data not shown). Thus, changes in hydrophobicity cannot account for
the differences in stop-transfer activity observed for the various STE sequences.
Integration of Transmembrane Proteins--
To determine if the
transmembrane molecules actually integrated in the ER membrane or if
they are translocation intermediates that are "stuck" in the
translocon, membrane integration was assayed by the rigorous criterion
of resistance to extraction from membranes by incubation in 1 M urea, 0.2 M sodium carbonate (pH 11.5) (19) (Fig. 4F). As expected when using such harsh conditions, the
fraction of molecules scored as integrated is in general slightly less than the number determined to be transmembrane by proteolysis. Nevertheless, the correlation between the number of negative charges and membrane integration (slope = 8.3; r = 0.79)
is still striking. Since, non-targeted molecules would complicate
quantification of carbonate extraction data, gel filtration
chromatography was used to remove non-targeted molecules prior to
incubation in 1 M urea, 0.2 M sodium carbonate
(pH 11.5).
In contrast to proteolysis assays, type I molecules cannot be
distinguished from type II molecules by carbonate extraction. To
confirm that both type I and type II transmembrane molecules were
integrated into the ER membrane, we labeled the Cys residues in type I
molecules with GMBSA prior to carbonate extraction. Unlike other
Cys-modifying reagents, GMBSA does not cross ER
membranes.2 Reaction with
GMBSA results in covalent attachment of the reagent to Cys residues,
thereby adding 3 positive charges (but only approximately 300 in
molecular weight) to the polypeptide for each modified Cys. Although
not all Cys residues are equally accessible to the labeling reagent,
sufficient labeling occurs at the 4 Cys residues in the Pt domain, to
permit labeled type I molecules to be differentiated from unlabeled
type II molecules. As controls we labeled the 
-subunit of signal
recognition particle receptor (a type I transmembrane protein
containing three Cys residues) and showed that the 6 Cys residues in
prolactin were protected from labeling with GMBSA after translocation
into the ER (data not shown).
By using this technique, even type I molecules that make up only 10%
of the total can be identified and shown to be resistant to extraction
with sodium carbonate (Fig.
5A, compare lanes 3 and 6, downward-pointing arrowheads, with lanes 2 and 5, respectively). The upward pointing
arrowheads indicate molecules integrated in the type II
orientation that were not modified with GMBSA. Signal cleaved molecules
(upward arrows) are protected from modification with GMBSA
but are not resistant to extraction with sodium carbonate because they
are located in the lumen of the ER.
View larger version (33K):
[in this window]
[in a new window]
|
Fig. 5.
Identification of transmembrane molecules
with type I topology by chemical labeling. A, GMBSA labeling
before carbonate extraction demonstrated that both type I and type II
polypeptides were integrated into the ER membrane. Polypeptides
containing the A2 or A8 STE, as indicated above the
panels, were synthesized in reticulocyte lysate containing
microsomes and then the microsomes and associated proteins were
isolated by gel filtration chromatography (lanes 1 and
4). An aliquot of the excluded fraction was labeled with
GMBSA, incubated in sodium carbonate (pH 11.5), and integral membrane
proteins (lanes 3 and 6) were pelleted by
centrifugation (P), whereas secreted or peripheral proteins
(lanes 2 and 5) remained in the supernatant
(S). Upward arrows, signal-cleaved secreted
molecules; upward arrowheads, type II integral membrane
proteins; downward arrowheads, type I integral membrane
proteins. B, the fraction of transmembrane molecules with
type I topology as determined by GMBSA labeling (squares) is
similar to that calculated from proteolysis data
(triangles).
|
|
Only a small fraction of the molecules with the A2 STE pellet after
carbonate extraction (Fig. 5A, compare the signal cleaved molecules indicated by the upward arrow in lane 2 with the corresponding band in lane 3). However, more than
half of the pelleted molecules are integrated with type I topology
because they are modified with GMBSA (downward pointing
arrowhead in lane 3). In contrast, the A8 STE
efficiently stops translocation, and molecules with either type I
(downward pointing arrowhead, lane 6) or type II topology
(upward pointing arrowhead, lane 6) were integrated in the
ER membrane. Extending this analysis to other STE sequences revealed
that in all cases the majority of type I membrane-associated molecules
were carbonate-resistant.
GMBSA labeling also permitted us to measure the fraction of each
molecule that adopted the type I topology. After GMBSA labeling the
fraction of molecules adopting type I topology can be measured directly
from the PhosphorImager plate rather than by calculation from
proteolysis data. Direct comparison of the results obtained using both
methods revealed excellent agreement. Thus, our estimates for the
fraction of molecules that adopted type I topology are reliable (Fig.
5B).
Probing the Environment of the STE Sequence during
Integration--
To examine the mechanism of STE-mediated membrane
integration in more detail, we used photocross-linking (26, 31). For these experiments, nascent chains containing the STE, hydrophobic region, and 51 amino acids of the Pt passenger were generated by
in vitro translation of truncated mRNAs. Truncation of
the nascent chain at amino acid 51 of Pt (by restriction digestion of
the plasmid with TaqI) positions the amino-terminal end
within the translocon but not far enough to permit signal peptide
cleavage. This allowed us to compare the immediate environment of both
functional and non-functional STE sequences within the translocon by
photocross-linking. To identify cross-links dependent on proximity to a
functional STE, we compared the photoadducts obtained for nascent
chains with the A13, A12, and A2 STEs. The only difference between the A12 and A13 STEs is the presence of a lysine in place of asparagine at
amino acid 11 of STE A12. Since these STEs function similarly (70-75%
transmembrane molecules), any photocross-links observed for STE A12,
but not with STE A13, are due to the additional probe inserted in the
center of the A12 STE. Photoadducts formed in both A12 and A13 nascent
polypeptide samples result from probes positioned within the coding
region of the mature portion of the Pt passenger (amino acids 12 and 49 of Pt). Since the truncation point is at amino acid 51 of Pt, residue
12 is predicted to have exited from the ribosome while residue 49 will
be within the ribosome.
The centrally located lysine in STE A12 is also present in STE A2, but
5 of the 6 glutamic acids in A12 were replaced with uncharged residues
(Fig. 1). Since less than 30% of the nascent polypeptides with STE A2
end up transmembrane, any photocross-links that require a functional
STE should be reduced (but not completely eliminated) for these
molecules when compared with photocross-links to nascent polypeptides
containing STE A12.
After photolysis, the translation reactions were fractionated by
carbonate extraction or by affinity chromatography and analyzed by
SDS-PAGE. When samples were pelleted after incubation in sodium carbonate (pH 11.5), the major cross-linked species migrated as a pair
of bands that run just above and below the 43-kDa marker. These species
were observed with all three nascent polypeptides, and hence they were
not dependent on the lysine probe located within the STE. However, a
cross-linked species that migrates between the 68- and 97-kDa markers
was prominent only for nascent chains that contained the A12 STE (Fig.
6, lane 2, downward arrow). The appearance of this photoadduct correlates with STE function since
the A12 STE is both functional and contains the centrally located
lysine residue. Subtracting the apparent molecular mass of the nascent
polypeptide from that of this photoadduct yields a predicted molecular
mass of 69 ± 6 kDa for the photocross-linking target.
View larger version (80K):
[in this window]
[in a new window]
|
Fig. 6.
Cross-linking with
N-(5-azido-2-nitrobenzoyl)-Lys-tRNA reveals
STE-specific cross-links. Translocation intermediates were
synthesized by translation of truncated transcripts in wheat germ
extract containing ER microsomes and the signal recognition particle.
After translation for 1 h, the samples were placed on ice and
photolyzed. Aliquots of the translation reactions were then
fractionated by incubation in sodium carbonate (pH 11.5) and pelleting
of integral membrane proteins by centrifugation, by passage through
concanavalin A-Sepharose to bind glycoproteins, or by
immunoprecipitation with antisera to TRAM. The A13 STE contains 6 Glu
residues, but no lysine, and hence serves as a control for cross-links
formed via photoreactive-modified lysine residues in the Pt reporter
protein. The A12 STE is similar but contains an Asn to Lys substitution
at position 11. Greater than 70% of A12 and A13 molecules adopt a
transmembrane topology. The A2 STE contains 1 Glu and 1 Lys, and less
than 30% of these molecules become transmembrane. The
downward-pointing arrow in lane 2 indicates a
photoadduct formed between a photoreactive probe on the modified lysine
in the A12 STE and a 69-kDa species. The downward-pointing
and upward-pointing arrows in lane 6 indicate
photoadducts formed between the photoreactive probe in the A12 STE and
glycoproteins of 54 and 34 kDa, respectively. The migration of
molecular mass markers is indicated (in kDa) to the left of
the figure.
|
|
When binding to concanavalin A-Sepharose was used to concentrate and
detect glycoprotein-containing photoadducts generated with the
non-glycosylated nascent polypeptides, no photoadduct with a 69-kDa
photocross-linking target was observed, suggesting that this
photocross-linking target is not glycosylated. Instead, a new
cross-link was visualized with STE A12 (Fig. 6, lane 5, downward
arrow). In addition, a band observed just below the 43-kDa marker
is darker in lane 5 than in lanes 4 or
6 (upward arrow). Subtracting the molecular mass
of the nascent polypeptide from the apparent molecular mass of these
species suggests that these bands originate from STE cross-links to
glycoproteins of approximately 54 and 34 kDa, respectively. Some
cross-linking to the 34-kDa molecule was also obtained with nascent
chains containing the A13 STE. This STE is functional but does not
contain a lysine within the STE sequence. Thus, it seems likely that
the 34-kDa glycoprotein can be cross-linked to nascent polypeptides
with the A13 STE by the lysine at amino acid 12 of the Pt passenger. The greater extent of photocross-linking to the 34-kDa target observed
in lane 5 than in lane 4 suggests that the 34-kDa
target can also react covalently with the probe within the A12 STE. In addition, cross-linking to species with similar migration was observed
in the carbonate pellets (lanes 1 and 2)
suggesting that the 34-kDa molecule is a membrane glycoprotein, most
likely TRAM (see below). In contrast, the 54-kDa target appears to be
photocross-linked solely via the STE probe, since neither A13 nor A2
reacted significantly with the 54-kDa glycoprotein (Fig. 6, compare
lane 5 with lanes 4 and 6).
Immunoprecipitation with antibodies to TRAM revealed very little
cross-linking to TRAM of nascent polypeptides with the A13 STE,
suggesting that the lysine residues in the Pt passenger domain of these
molecules are not adjacent to TRAM. In contrast, both STE A12 and the
poorly functional STE A2 are in close enough proximity to TRAM to
permit cross-linking (Fig. 6, lanes 7-9). Whereas
cross-linking of TRAM to nascent polypeptides with STE A12 reproducibly
resulted in a single sharp band, cross-linking of TRAM to A2 was always diffuse (Fig. 6, compare lane 8 with lane 9). Yet
the total yield of the STE A2 photoadduct was roughly equal that of the
STE A12 photoadduct. Finally, no STE-dependent
photocross-links to Sec61
were observed using antiserum to Sec61,
and photocross-linking to Sec61 was very inefficient for all three
nascent polypeptides (data not shown). The extent of cross-linking to
Sec61 differs for probes at different positions within a nascent
polypeptide (32, 33). Thus, It seems likely that the lack of
cross-linking to Sec61 observed here is due to the position of the
probes within the STEs.
| |
DISCUSSION |
The behavior of the STE sequences examined here is unprecedented.
Current models presume that hydrophobicity is the sole or primary
determinant of integration into the ER membrane, yet the negatively
charged Glu residues in the IgM STE sequence clearly affect the extent
of protein integration for hydrophobic sequences of as few as 15 residues. How does a functional STE cause the integration of marginally
hydrophobic sequences that would otherwise be translocated across the
ER membrane? Do specific proteins mediate the STE effect?
Although we cannot determine where in the ER membrane the STE is
recognized, our cross-linking data suggest that it occurs within the translocon.
Our photocross-linking experiments have identified three proteins that
are adjacent to the STE sequence in the translocon. One of these
proteins is TRAM, a previously identified component of the translocon.
But the other two proteins, a 69-kDa membrane protein and a 54-kDa
glycoprotein, do not resemble any proteins identified previously in
studies with nascent secretory or membrane proteins (26, 27, 31-33).
These results therefore suggest that STE-mediated integration into
membranes may be affected by specific proteins, a conclusion that is
consistent with the relative insensitivity of this effect to the
hydrophobicity of the nascent chain.
Surprisingly, the effect of the Glu residues on conversion of a signal
sequence to a signal-anchor sequence was additive. The additive effect
of the negatively charged residues on membrane integration suggests
that, similar to the effects of positively charged amino acids on
membrane topology (34), the negative charges in the STE contribute to a
feature required to trigger membrane integration. This result, obtained
in two different eukaryotic cell-free systems and for two different
hydrophobic sequences, is in direct contrast with results obtained
using E. coli in which there was no effect of amino-terminal
negative charges on signal sequences integration into membranes (28).
Furthermore, although negative charges at the amino terminus of a
polypeptide have been predicted to cause the polypeptide to adopt a
type I topology (35-37), the cumulative effect of the negative charges
was primarily to increase the fraction of molecules that integrated
into the bilayer and adopted the type II topology. The integration of
these proteins with a topology opposite to that predicted by rules
based on the distribution of charges surrounding a transmembrane
sequence (2, 5) is consistent with the possibility that STE-mediated integration into membranes occurs via a mechanism distinct from integration of conventional signal-anchor sequences.
As expected for true integral membrane proteins, polypeptides that
adopted a transmembrane topology due to a functional STE sequence were
not only resistant to extraction with sodium carbonate (pH 11.5) (Fig.
5B) but were also resistant to extraction with 2 M urea, 1 M NaSCN, or 0.08% (w/v) sodium
deoxycholate (data not shown). Therefore, transmembrane orientation
does not result from a simple STE-dependent steric block of
translocation. Consistent with this interpretation, the various STE
sequences also did not alter the targeting efficiency of the molecules
(data not shown). Significantly, the only other STE sequence
characterized thus far (that of PrP) was recently demonstrated to
integrate PrP molecules in the ER membrane with both type I and type II
topology (25). Thus, one characteristic of STE sequences may be that
they can eventually reside on either side of the ER membrane.
Furthermore, the IgM STE sequences that function at the amino terminus
also stop translocation of 40 and 70% of hydrophobic sequences
containing 15 or 17 uncharged residues, respectively, when located in
the middle of a translocating polypeptide (Fig. 2). We conclude
therefore, that the STE and hydrophobic sequences elicit membrane
integration no matter where they are located in the translocating polypeptide.
To analyze the IgM STE by proteolysis and chemical labeling, we used a
reticulocyte lysate cell-free system because it has been shown to
reflect more accurately the topology of transmembrane molecules in
transfected cells than does wheat germ extract (38). Yet while the
integration of most polypeptides into the ER membrane is qualitatively
similar in wheat germ extract, reticulocyte lysate, and in transfected
mammalian cells, the only other STE sequence identified to date behaves
differently in wheat germ extract and reticulocyte lysate. In wheat
germ extract, the PrP STE exhibits dramatically increased membrane
integration (20). Therefore, we also assayed all of the molecules in
Fig. 1 using this system. Unlike PrP, membrane integration of molecules
with the IgM STE and the mutant STEs examined here was quantitatively
similar in both systems (data not shown). Thus, it appears that for the
IgM STE, a membrane-dependent recognition event leads to
STE-mediated integration of the nascent polypeptide into the membrane.
The nature of this recognition event is unknown, but our
photocross-linking experiments showed that when a functional STE enters
the translocon, it is proximal to different proteins than either the
mature portion of the nascent chain or a non-functional STE (Fig. 6).
The simplest explanation for these data is that recognition of a
functional STE moves the nascent chain to a distinct region within the translocon.
However, this transmembrane domain has been shown to interact with
several different proteins in B lymphocytes and immature B cells (39,
40). Although one of these has a molecular mass of 32 kDa (prohibitin),
it is not known if there is a counterpart in canine pancreatic
microsomes. Furthermore, prohibitin is not a glycoprotein. Thus, it
seems likely that two of the cross-linking partners identified here are
novel ER proteins. It remains to be determined whether the same
polypeptides will be found adjacent to other STE sequences.
At present it is not possible to determine how many proteins may be
integrated in the ER membrane via STE sequences because there is no
consensus sequence information, and the limited hydrophobicity of
potential transmembrane sequences may have allowed them to escape
detection. It is also possible that STEs may contribute to membrane
integration of proteins that contain a hydrophobic sequence that is
sufficiently hydrophobic to mediate integration independent of an STE
sequence (such as transmembrane immunoglobulins). In these cases we
speculate that, similar to gene products that appear to be functionally
redundant, STE sequences may be important for more subtle
characteristics of the integrated species or may be vestiges of
evolution that are no longer essential. Consistent with the former, the
IgM transmembrane domain has been shown to acquire a post-translational
modification that increases the hydrophobicity of the sequence (41).
This modification cannot account for the results obtained here with the
various STE sequences as the modification does not occur in our
cell-free system (data not shown).
Taken together, our results demonstrate that co-translational
integration of proteins into the ER membrane is not always regulated strictly by hydrophobicity, and suggest that current models for membrane integration should be modified to include a mechanism for
regulating integration. Indeed our demonstration that a hydrophobic sequence of 23 uncharged amino acids functioned as an efficient signal
peptide, while many constructs containing only 17 uncharged amino acids
were integrated into the ER membrane, suggests that predictive
algorithms based only on hydrophobicity provide an incomplete view of
membrane topology.
,
TOP
ABSTRACT
INTRODUCTION
